With brand-new systemwide Writing Tools powered by Apple Intelligence, users can rewrite, proofread and summarise text nearly everywhere they write.
Apple Intelligence is available from today in New Zealand for those with a recent iPhone, iPad or Mac (see compatibility details at the foot of this story).
There are scads of new features, including a smarter, more patient Siri - now denoted by a glow around the edge of the screen - with the voice assistant staying active and coping with stumbles and rewording, rather than you having to carefully time a precise question.
In a nice discrete touch, you can double-tap the bottom of the screen to ask Siri a question via text. The new Siri can also answer thousands of questions about Apple products, such as “How do I share a wi-fi password?”
More complex queries can be thrown to OpenAI’s ChatGPT-4o, which is integrated into Apple Intelligence (the company says other generative AIs will be integrated over time).
You get the option to access ChatGPT 4o for free without being logged in, and Apple obscures your IP address for an extra layer of anonymity and ChatGPT won’t store requests received via Apple Intelligence.
That’s part of Apple Intelligence’s central focus on privacy. Measures include:
As many queries as possible are processed on your device as possible (a flip of the usual AI paradigm)
Queries that can’t be handled on your device are handled by an Apple Private Cloud Compute-capable data centre − which the company says is essentially a giant extension of your device, with all of the same privacy protection applying.
There are new options to clean up photos (more on this below), image generation from casual sketches with Image Playground and AI smarts to prioritise your inbox and summarise messages − including searching for key tasks in tangled back-and-forth threads.
Voice and FaceTime calls can now be recorded. An audio file and transcription are saved to Notes. Apple Intelligence generates a summary to help recall key points.
I thought New Zealand might miss out on the option to record phone calls - but I was able to record a call this morning using the new feature this morning between two mobile numbers. The person you’re calling hears Siri saying, “This call will be recorded” if you choose to record a call.
There are new options to summarise or smarten up writing across multiple apps with professional, concise, or friendly tones, plus the option to abbreviate a slab of text or change it to bullet points − to, in the other direction, convert a document to a. The general approach is for Apple Intelligence features to appear where you need them rather than a dedicated app.
You can create a “Genmoji” by typing a description into the emoji search box.
And you can search your Photos library with a natural language query. I tried “Our ginger and black cats together” and got an instant album of said moggies. The feature works across videos, too.
A long press on an iPhone 16’s camera button brings up a new Visual Intelligence search option, letting you query Google about a restaurant or product or anything else in your camera’s view. Think Google’s Circle to Search on Android but without the hassle of having to circle something.
Visual intelligence can also be used to summarise and copy text, translate text between languages, detect phone numbers or email addresses with the option to add to contacts and more. Camera Control also allows users to search Google so they can see where they can buy an item, or benefit from ChatGPT’s problem-solving skills
Rivals are offering similar AI features, even if fans will argue they’re more user-friendly and intuitive in Apple Intelligence. But perhaps the signature feature is the firm’s focus on privacy.
“Unlike other Big Tech companies, Apple has really leaned into a pro-privacy position and embedded consumer trust as a core part of its brand,’ says Frith Tweedie, an intellectual property lawyer who now runs an AI-focussed privacy consultancy and is an executive council member of the NZ AI Forum.
“While that’s easier for Apple to do given it’s got a much less data-hungry business model than say Meta or Google, generally speaking, Apple does tend to put its money where its mouth is.”
Apple Intelligence certainly looks promising from a privacy perspective. Most processing of personal information is done on your device, with more complex tasks carried out using Apple’s Private Cloud Compute, whose models don’t store personal information. Plus those servers are publicly available so independent researchers can verify Apple’s claims - a promising transparency move.
Finally, Apple allows users to choose to use ChatGPT, rather than automatically opting them in. And even if you turn on the ChatGPT option, you’re still asked each time if you’re sure you want to use it. Again, great transparency and real user choice, both of which support privacy and help drive trust.
“One of the things we know from our customers and people around the globe is that they are very concerned about what’s happening with their data and their information, especially in the context of AI,” Apple VP of worldwide product marketing Bob Borchers told the Herald.
“Apple is taking a very different approach to AI than other Big Tech companies. By building their own silicon, they have integrated their own AI chips directly into their new phones, iPads and laptops,” says Victoria University senior lecturer in artificial intelligence Andrew Lensen.
“This gives them a lot of control over their use of AI, allowing for many AI requests to be performed on the user’s own device. This is very different to other approaches, such as those using ChatGPT, where your data must be sent to an external server every time you use an AI function. This is a big part of Apple’s privacy advantage: if data doesn’t leave your machine, it is much easier for your privacy to be protected.
“Despite their progress in developing AI chips, they still say that complex AI requests will need to be sent to a “Private Cloud Compute”, which is a server in an Apple data centre.
“They have promised that only the necessary data will be sent and that these servers are carefully designed to guarantee privacy. This is an inevitable compromise – many uses of AI are highly resource-intensive and cannot be run on your own machine.
“So, we have to trust Apple’s word for these usages − much like with other AI tools. Still, Apple’s hybrid approach and their slower AI rollout do make them feel more trustworthy than companies like Microsoft, who threw caution to the wind to be AI-first.”
Microsoft chief cybersecurity advisor Bret Arsenault told the benchmarks no AI product was released unless it met regulatory requirements and strict in-house security and governance benchmarks.
“It’s very easy to treat generative AI tools as a plaything or harmless tool. Using models like ChatGPT to automate tasks is incredibly convenient and often a big time-saver,” Lensen says.
“But what information are you giving to any of these models? Data is the currency of a Generative AI world, and companies with a data edge can develop the best AI models. These companies could use your data to train their next model, sell your data for marketing, or, if hacked, your sensitive data could be used to blackmail you.”
“Even if the privacy policy says Apple will do none of this – and they have groundbreaking encryption – all the existing cybersecurity issues, man-in-the-middle attacks and so on, are still prevalent.”
Borchers says data was fully encrypted in transit and at either end. The all-Apple technology approach allowed it to wrap security tightly on-device and in Private Cloud Compute where “Apple silicon servers never store your data. They provide the response, then everything is removed.”
It’s also a case of his company literally putting its money where its mount is. Apple has offered a US$1 million ($1.7m) bounty to any “white hat” hacker who can expose a flaw in Private Cloud Compute.
Borchers sees it as a measure to help build public trust.
More broadly, Apple has made resources, including a virtual research environment, available to “all security and privacy researchers − or anyone with interest and a technical curiosity − can hold us accountable to our privacy promises,” Borchers says.
“This certainly throws down the gauntlet to Apple’s main competitor Google to come up with similar privacy-guaranteeing service levels for Gemini services. Right now we must assume that everything typed into any other AI cloud service from Google, OpenAI, Anthropic, Meta or Mistral is surveillable. But does the market care enough?” futurist Ben Reid says.
Reid saw AI tools from Google, Anthropic and AI with a performance advantage at this point.
“Will Apple Intelligence performance catch up with leading AI labs like Google, Anthropic and OpenAI?” he asks.
“Or will AI privacy remain a tradeoff against performance into the future? If Apple does catch up in performance, then I think we’ll see other labs being forced to implement similar privacy architectures to remain competitive.”
Notwithstanding that ChatGCP can currently respond to a broader range of queries, Borchers says Apple in fact had a “distinct performance advantage” in that, “We’ve spent more than a decade developing Apple Silicon and a specialised OS to run it ... The more we do on-device, the faster the performance and the lower the energy cost.”
Reid adds, “If you think forward a few years from now, with AI becoming far more deeply integrated into our personal thinking processes than it is now − it will effectively become an augmentation to the human mind. At that point, it feels likely that many people would want both a privacy guarantee and open source technology guarantee for platforms which manage their own thoughts.”
Open source is publically shared code.
“I think the premise that we should be, however people are providing AI should be, transparent and accessible and audited to the degree that somebody can hold you accountable to your promises,” Borchers says.
Reid’s mind-meld is not on the immediate horizon, but Apple says, “Additional Apple Intelligence capabilities will be available in the months to come. Siri will be even more capable, with the ability to draw on a user’s personal context to deliver intelligence that’s tailored to them.
“Siri will also gain onscreen awareness and will be able to take hundreds of new actions in and across Apple and third-party apps. Priority notifications will also surface what’s most important. In addition, users will be able to create images in Image Playground in a Sketch style, an academic and highly detailed style that uses a vibrant colour palette combined with technical lines to produce realistic drawings.”
Apple Intelligence is available now as a free software update with iOS 18.2, iPadOS 18.2 and macOS Sequoia 15.2, and can be accessed in most regions around the world when the device and Siri language are set to localised, English for Australia, Canada, Ireland, New Zealand, South Africa, the UK or the US.
Apple Intelligence is available on iPhone 16, iPhone 16 Plus, iPhone 16 Pro, iPhone 16 Pro Max, iPhone 15 Pro, iPhone 15 Pro Max, iPad with A17 Pro or M1 and later, and Mac with M1 and later.
Chris Keall is an Auckland-based member of the Herald’s business team. He joined the Herald in 2018 and is the technology editor and a senior business writer.
Financial Times: Now comes the challenge of keeping them in line.